Elsevier

Cognitive Psychology

Volume 92, February 2017, Pages 127-140
Cognitive Psychology

Habit outweighs planning in grasp selection for object manipulation

https://doi.org/10.1016/j.cogpsych.2016.11.008Get rights and content

Highlights

  • Grasp are adjusted to the requirements of upcoming object manipulation tasks.

  • Previous (but obsolete) task requirements mostly determine grasp adjustments.

  • Expected task requirements have a comparatively small effect on grasp adjustments.

  • Thus, a hybrid habitual and goal-directed process determines grasp adjustments.

Abstract

Object-directed grasping movements are adapted to intended interactions with an object. We address whether adjusting the grasp for object manipulation is controlled habitually, based on past experiences, or by goal-directed planning, based on an evaluation of the expected action outcomes. Therefore, we asked participants to grasp and rotate a dial. In such tasks, participants typically grasp the dial with an excursed, uncomfortable arm posture, which then allows to complete the dial rotation in a comfortable end-state. We extended this task by manipulating the contingency between the orientation of the grasp and the resulting end-state of the arm. A one-step (control) group rotated the dial to a single target. A two-step group rotated the dial to an initial target and then in the opposite direction. A three-step group rotated the dial to the initial target, then in the opposite direction, and then back to the initial target. During practice, the two-step and three-step groups reduced the excursion of their grasps, thus avoiding overly excursed arm postures after the second rotation. When the two-step and three-step groups were asked to execute one-step rotations, their grasps resembled those that were acquired during the two-step and three-step rotations, respectively. However, the carry-over was not complete. This suggests that adjusting grasps for forthcoming object manipulations is controlled by a mixture of habitual and goal-directed processes. In the present experiment, the former contributed approximately twice as much to grasp selection than the latter.

Introduction

When we plan an action, we often have subsequent actions in mind. This becomes evident as the way we execute initial actions depends on the actions that follow (Ansuini et al., 2006, Cohen and Rosenbaum, 2004, Gentilucci et al., 1997, Marteniuk et al., 1987, Rosenbaum et al., 1990, Sartori et al., 2011). Such anticipatory behavior is particularly important for grasping a to-be-manipulated object because most object manipulations are best executed with a particular grasp. For example, a person who wants to rotate a door-knob in a clockwise direction will rotate the arm counterclockwise before grasping it and vice versa. This maintains the arm posture in a neutral medial range during the object manipulation and increases its’ speed and accuracy (Herbort, 2015, Rosenbaum et al., 1996, Short and Cauraugh, 1999).

Before an object can be grasped, the grasping movement must be planned. This process includes several aspects, such as specifying the direction of the movement, shaping the fingers, or determining the force with which the object will be grasped. Here, we focus on the following specific – but central – aspect of this planning process: selecting how to place the fingers on an object based on the intended object manipulation (“grasp selection for object manipulation”). Although grasp selection for object manipulation has extensively been studied (for recent reviews, see Rosenbaum et al., 2012, Wunsch et al., 2013), little is known about the underlying mechanisms. There are two different perspectives that correspond to a dichotomy between goal-directed and habitual action selection (Dickinson, 1985, Dolan and Dayan, 2013). According to one approach, grasp selection for object manipulation is primarily a goal-directed planning process that is based on the anticipated action outcomes (Cohen and Rosenbaum, 2004, Johnson, 2000, Wunsch and Weigelt, 2016). This notion is goal-directed as grasp selection depends on the expected consequences of a grasp (e.g., the resulting arm posture) and matching these consequences to an individual’s motivations (e.g., assuming a comfortable posture). According to the other approach, grasp selection is primarily habitual and is based on learned object manipulation task – grasp associations (Herbort et al., 2014, van Swieten et al., 2010). This is habitual because it assumes that grasps are selected because they proved useful in the past for manipulating objects in comparable ways, regardless of the expected requirements for the upcoming task. However, there is no compelling evidence for either perspective. In the remainder of the introduction, we present arguments for both views and outline the experimental procedure used to test between them.

According to the goal-directed view of grasp selection, anticipating the arm movement that is necessary to manipulate an object is used to select a grasp that allows for fast, accurate, or comfortable object manipulations (Cohen and Rosenbaum, 2004, Johnson, 2000, Stöckel et al., 2012, Wunsch and Weigelt, 2016). Thereby, the arm posture at the end of the object manipulation (end posture) and the arm posture when the object is grasped (initial posture) seem to play a pivotal role. Notably, this view includes the possibility that planning may only be necessary when a grasp is selected for a specific task for the first time. When a task is repeated, grasp selections may rely on recalling previous instances (Cohen and Rosenbaum, 2004, Weigelt et al., 2007).

There are mainly two observations that support the goal-directed view; however, they are not conclusive per se. First, at least some of the cognitive requisites for selecting grasps based on anticipated end postures are met. For example, to plan grasps that are based on the resulting end-states, it is necessary to prospectively predict and evaluate the possible end-states. Indeed, participants can mentally simulate object manipulations (Seegelke & Hughes, 2015). Likewise, participants can predict the subjective “awkwardness” of the arm postures (Johnson, 2000), which are a key determinant of grasp selection (Rosenbaum, Vaughan, Barnes, & Jorgensen, 1992). Moreover, movement end postures appear to be represented prior to the onset of movement. For example, prospective judgments of how an object could be grasped for rotation were faster when the participant’s actual arm posture was congruent to the arm posture at the end of the object manipulation (Zimmermann, Meulenbroek, & de Lange, 2012). Finally, the ability to discriminate between visual images of comfortable and awkward postures is correlated with the ability to adapt grasps for different object manipulations (Stöckel et al., 2012). However, the representation of the end postures does not imply that this information is processed during planning or that planning occurs at all (Johnson, 2000). These representations could be the result, rather than the cause, of action selection (Blakemore et al., 2002, Ziessler and Nattkemper, 2011).

Second, grasp selection often depends on the intended object manipulation from the very first trial on which the task is performed (e.g., Cohen and Rosenbaum, 2004, Rosenbaum et al., 1990). As there is no opportunity for learning, these experiment suggests that grasps are planned in goal-directed fashion (Cohen & Rosenbaum, 2004). However, because most experimental tasks are inspired by everyday actions (Rosenbaum et al., 2012), participants may have reused the task-grasp associations that were learned during daily object manipulations. In fact, in less common tasks, participants made little to no grasp adjustments for different object manipulations on the very first trial(s) and only adjusted their grasps after gaining some experience with the task (Herbort, 2012, Künzell et al., 2013).

There are (at least) two ways that grasp selection could depend on habitual processes. First, specific grasps may be associated with different objects, regardless of the task. As such, specific grasps may not reflect the currently intended interaction with that object. For example, humans who are asked to manipulate everyday objects tend to select grasp points or grasp orientations that correspond to the object’s prevailing use, regardless of their current object manipulation goals (Creem and Proffitt, 2001, Herbort and Butz, 2011). Likewise, grasps tend to be conserved during repeated (identical) interaction with objects (Glover and Dixon, 2013, Rosenbaum and Jorgensen, 1992). In this case, habitual processing thwarts adjusting grasps to object manipulations to some extent.

Second, specific grasps may be associated with specific object manipulations (i.e., a combination of object and intended object manipulations). The present article focuses on this aspect, which we refer to as the habitual view. This view suggests that specific grasps are selected for specific object manipulations because they have previously been used successfully for similar object manipulations (Herbort et al., 2014, van Swieten et al., 2010). Hence, grasp selections may depend on intended object manipulations because participants recollect grasps that were previously the most suitable for manipulating an object in a specific way (and not because participants anticipate, for example, the end-states that resulted from possible grasps). According to this perspective, habitual processing is the primary cause of adjusting grasps to object manipulations.

This approach may be considered goal-oriented because it assumes a certain intended goal state prior to executing an action. However, it is habitual because it assumes that these goal states invoke “dumb” recollections of actions that were previously helpful for reaching the same end goal, rather than assuming “clever” planning processes that bridge current states and intended end states. Although it may appear counterintuitive, goal-directed control processes have been shown to trigger one or more actions that were under habitual control (Dayan, 2009, Dezfouli and Balleine, 2013, Dezfouli et al., 2014). For example, the goal-directed decision to cycle to the office or go to the supermarket will trigger different action sequences. However, the individual actions in the sequences (e.g., taking specific turns at intersections depending on the ultimate goal) might be driven by habit because the route that is chosen is based on past experiences and not on the evaluation of the outcomes that are associated with each individual option. Likewise, the habitual view posits that the intent to conduct an object manipulation may trigger a specific habitually controlled grasp selection.

To our knowledge, there is no direct evidence for the habitual view, but these processes may theoretically account for anticipatory grasp selection. For example, one computational model of grasp selection for object manipulation in continuous tasks accurately explains the empirical data without predicting the potential action outcomes (Herbort, 2013, Herbort and Butz, 2012). Hence, goal-directed planning does not seem necessary to explain grasp selection for object manipulation.

In summary, it is not possible to determine how a grasp is selected when planning a grasping movement because there are no conclusive findings for the goal-directed or habitual approaches. Thus, in this article, we directly compare both approaches. We focus on the contributions of goal-directed or habitual processes on grasp selection, without claiming that other parts of movement planning will equally be habitual or goal-directed.

One way to test these hypotheses is to create a situation in which participants’ expectancies about the consequences of selecting a specific grasp in the next trial differ from the actual consequences that were experienced in previous trials. If grasp selection is goal-directed, grasps should depend on the expected relationship between grasps and their consequences. That is, if participants know the changed relationship between the grasps and their consequences, they should immediately or quickly adapt their grasps to the changed consequences. In contrast, current grasps should depend on the consequences of previous grasp selections if they are under habitual control. That is, even when participants know the changed relationship between the grasps and their consequences, they should continue to use the grasps that were suitable in previous trials.

The present experiment is based on this rationale. We manipulated the consequences of grasp selection with a multi-step object manipulation task. Participants were asked to grasp a knob that was attached to a pointer, and rotate it to various targets (Fig. 1a). Three independent groups of participants experienced different relationships between their grasp selections and action outcomes during learning blocks. In these blocks, the one-step group rotated the pointer to a target and then released the dial. The two-step group rotated the dial to a target and then immediately rotated the pointer to a target that was in the exact opposite position without releasing the dial between both rotation steps (two-step rotations; c.f. Fig. 1b). Thus, grasps that resulted in a comfortable posture when reaching the first target ultimately resulted in an uncomfortable end posture. The three-step group rotated the dial to the target, then immediately to a target at the opposite position, and then immediately back to the position of the initial target (three-step rotations). Because each initial target was followed by a specific second (and third) target, the target sequences were fully predictable (c.f. Fig. 1b). Importantly, participants benefit less from adapting their grasp to the initial rotation segment in the two-step and three-step rotations because a comfortable posture after the first (and third) rotation implies that there is an uncomfortable posture after the second rotation. Hence, it is expected that grasp selection will change during the learning blocks for the two- and three-step groups.

Although the three groups differed on the required learning block tasks, all participants executed one-step rotations from time to time in transfer blocks. The central question is how participants grasp the dial in the transfer blocks. If grasp selection is under habitual control, the grasps that were selected for rotation with a specific initial target in the preceding learning blocks should be used for one-step rotations with the same initial target in a subsequent transfer block. Because grasp selections are expected to differ between the one-step and the other groups in learning blocks, the habitual hypotheses suggest that grasp selections will also differ between the groups in transfer blocks. In contrast, there should be no difference between grasp selections for learning trials and transfer trials in each group.

If grasp selection is goal-directed, the grasps that are selected for the one-step rotations in the transfer blocks should not differ between groups because the relationship between the grasps and their consequences are the same in the transfer blocks regardless of group. In contrast, the grasps that are selected for rotations with a specific initial target in the learning blocks should differ from the grasps for rotations with the same initial target in the transfer blocks for the two- and three-step groups. In the one-step group, the learning and transfer block trials do not differ; thus, there should be no differences.1

Several measures were taken to ensure that goal-directed planning was possible in the transfer blocks, in which participants only executed one-step rotations. First, transfer trials were administered blockwise, and the type of block was announced prior to the start of the block (one-step vs. two-step vs. three-step). Hence, it is unlikely that participants did not register the task for the upcoming trial. Second, one-step rotations were used in transfer blocks because participants were expected to be most familiar with this task (and the data will suggest that this was the case). Thus, if participants engaged in goal-directed planning, they didn’t have to acquire a new mapping between grasps and their consequences to be able to plan the critical transfer trials. Third, the two- and three-step rotations always contained the one-step rotation as the initial part of the overall object manipulation, which allowed participants to maintain previously acquired knowledge from the one-step rotations during the learning blocks.

Section snippets

Participants

Thirty participants were recruited from the Würzburg area and were tested after providing informed consent (22 females, 8 males, mean age 27 years, sd = 7). All participants were right-handed (Lateral Preference Inventory, Coren, 1993).

Stimulus and apparatus

Fig. 1a shows the experimental layout. Participants were seated in front of a table, on which there was a start button and a dial. The start button (3 cm × 4 cm × 2.5 cm) was located between the participant and the dial (distance from dial: 35 cm). The dial consisted of a

Manipulation check: The effect of the learning condition on AGE in the learning blocks

The experiment hinges on the assumption that grasp selections differ between groups in learning blocks after the learning occurs. This assumption is asserted as follows. The connected lines in Fig. 2 show that the AGE did not change in the one-step group but decreased in the two- and three-step groups. A split-plot ANOVA2 on AGE with the within-subject factor of learning block and the between-subject factor of learning group

Discussion

We examined whether grasp selection for object manipulation is primarily based on goal-directed or habitual planning. To test our hypotheses, the postures that resulted from selecting a specific grasp were varied between three groups. In the one-step group, grasp selections did not change across four sessions, which suggests that the grasps were tuned to the tasks from the onset of the experiment. In the two-step and three-step groups, grasps were initially comparable to the one-step group but

Acknowledgments

This work was funded by Grant HE 6710/2-1 of the German Research Foundation (DFG) granted to Oliver Herbort. We thank Albrecht Sebald and Georg Schüssler for technical support.

References (52)

  • D.A. Rosenbaum et al.

    Planning macroscopic aspects of manual control

    Human Movement Science

    (1992)
  • C. Seegelke et al.

    The influence of action possibility and end-state comfort on motor imagery of manual action sequences

    Brain and Cognition

    (2015)
  • C. Seegelke et al.

    Grasp posture planning during multi-segment object manipulation tasks – Interaction between cognitive and biomechanical factors

    Acta Psychologica

    (2013)
  • R. Shadmehr

    Control of movements and temporal discounting of reward

    Current Opinion in Neurobiology

    (2010)
  • M.W. Short et al.

    Precision hypothesis and the end-state comfort effect

    Acta Psychologica

    (1999)
  • L. Wasserman

    Bayesian model selection and model averaging

    Journal of Mathematical Psychology

    (2000)
  • C. Ansuini et al.

    Effects of end-goal on hand shaping

    Journal of Neurophysiology

    (2006)
  • P.J. Bayley et al.

    Robust habit learning in the absence of awareness and independent of the medial temporal lobe

    Nature

    (2005)
  • C.J. Coelho et al.

    End-state comfort trumps handedness in object manipulation

    Journal of Experimental Psychology: Human Perception and Performance

    (2013)
  • R.G. Cohen et al.

    Where grasps are made reveals how grasps are planned: Generation and recall of motor plans

    Experimental Brain Research

    (2004)
  • R.G. Cohen et al.

    Prospective and retrospective effects in human motor control: Planning grasps for object rotation and translation

    Psychological Research

    (2011)
  • S. Coren

    The lateral preference inventory for measurement of handedness, footedness, eyedness, and earedness: Norms for young adults

    Bulletin of the Psychonomic Society

    (1993)
  • S.H. Creem et al.

    Grasping objects by their handles: A necessary interaction between cognition and action

    Journal of Experimental Psychology: Human Perception and Performance

    (2001)
  • A. Dezfouli et al.

    Actions, action sequences and habits: Evidence that goal-directed and habitual action control are hierarchically organized

    PLoS Computational Biology

    (2013)
  • A. Dezfouli et al.

    Habits as action sequences: Hierarchical action control and changes in outcome value

    Philosophical Transactions of the Royal Society of London B: Biological Sciences

    (2014)
  • A. Dickinson

    Actions and habits: The development of behavioural autonomy

    Philosophical Transactions of the Royal Society of London B, Biological Sciences

    (1985)
  • Cited by (15)

    • The ontogeny of efficient second-order action planning: The developing interplay of controlled and habitual processes in goal-directed actions

      2022, Journal of Experimental Child Psychology
      Citation Excerpt :

      Our results would then indicate that this anticipatory capacity improves over childhood (Wunsch & Weigelt, 2016). At the same time, however, recent studies with adults question whether grasp selections are solely based on expected action outcomes and highlight the role of past action experiences for grasp selection (Herbort, Kirsch, & Kunde, 2019; Herbort et al., 2017). This might explain why the relation is only moderate, suggesting an additional impact of previous task-specific adaptations.

    • The influence of time constraints on posture choices during an end-state comfort task

      2020, Human Movement Science
      Citation Excerpt :

      Several experiments have demonstrated that participants tend to persist in the use of the same grasps even when these grasps would have rated as more uncomfortable if not performed in sequence (Cohen & Rosenbaum, 2004, 2011; Weigelt et al., 2009). In addition, several studies have demonstrated a habitual effect in planning whereby grasps are planned based both on past experiences with grasping and moving objects, and on immediate task goals (Herbort, Mathew, & Kunde, 2017). Our main prediction, that individuals would shift toward beginning-state, rather than end-state comfort when the beginning-state was time constrained, was supported.

    • What's your number? The effects of trial order on the one-target advantage

      2018, Acta Psychologica
      Citation Excerpt :

      Hence, it may be that when one-target responses are repeated in a sequence, error (or variability) tolerance is heightened and movement times are reduced due to a speed accuracy tradeoff (Brenner & Smeets, 2011). Participants may have opted to use feedback from the previous trial with the advance information given about the up and coming trial (i.e., one- or two-segment movement: Herbort, Mathew, & Kunde, 2017). In single target movements, vision plays a role in adjusting movement programming from trial to trial (offline visual feedback processing) as well as during movement execution to correct errors in the limb trajectory (online visual feedback processing: Mackrous & Proteau, 2007).

    View all citing articles on Scopus
    View full text